helios/helios-remote
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

feat: wait-for-window command; add click guideline to SKILL.md

Browse source
This commit is contained in:
Helios Agent 2026-03-03 16:12:32 +01:00
parent 7ddaf5ddfe
commit e00270550d
No known key found for this signature in database
GPG key ID: C8259547CD8309B5
2 changed files with 42 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

17
skills/SKILL.md
View file

@ -80,9 +80,20 @@ python $SKILL_DIR/remote.py label "Moritz PC" "Neues Label"
4. `exec` / `click` / `type` → Aktion ausführen
5. `screenshot` → Ergebnis prüfen
## Wenn etwas geklickt werden muss das ich nicht finde
# Warten bis ein Fenster erscheint (z.B. nach Programmstart)
python $SKILL_DIR/remote.py wait-for-window "Moritz PC" "notepad" --timeout 10
```
## ⚠️ Klick-Regel (wichtig!)
**Niemals blind klicken.** Pixel-Koordinaten aus Screenshots sind unzuverlässig.
Wenn ich auf einen Button oder UI-Element klicken muss:
1. Erst `prompt` benutzen um Moritz zu bitten es selbst zu klicken
2. Dann weitermachen sobald er OK drückt
```bash
python $SKILL_DIR/remote.py prompt "Moritz PC" "Bitte klicke auf [X], dann drücke OK"
python $SKILL_DIR/remote.py prompt "Moritz PC" "Bitte klicke auf [Speichern], dann OK drücken"
```
→ Warte auf ACK, dann weitermachen.
Ausnahme: wenn ich die exakten Koordinaten eines Elements kenne (z.B. durch wiederholte Nutzung desselben UIs).

28
skills/remote.py
View file

@ -302,6 +302,28 @@ def cmd_prompt(args):
print(f"User confirmed prompt on session {sid!r}.")
def cmd_wait_for_window(args):
"""Poll until a visible window with the given title appears (or timeout)."""
import time
sid = resolve_session(args.session_id)
query = args.title.lower()
deadline = time.time() + args.timeout
interval = 1.0
print(f"Waiting for window matching '{args.title}' (timeout: {args.timeout}s)...")
while time.time() < deadline:
resp = _req("GET", f"/sessions/{sid}/windows")
windows = resp.json().get("windows", [])
matches = [w for w in windows if w.get("visible") and query in w.get("title", "").lower()]
if matches:
print(f"{'ID':<20} Title")
print("-" * 70)
for w in matches:
print(f"{str(w.get('id','?')):<20} {w.get('title','')}")
return
time.sleep(interval)
sys.exit(f"[helios-remote] Timeout: no window matching '{args.title}' appeared within {args.timeout}s")
def cmd_run(args):
"""Launch a program on the remote session (fire-and-forget)."""
sid = resolve_session(args.session_id)
@ -391,6 +413,12 @@ def build_parser() -> argparse.ArgumentParser:
fwp.add_argument("session_id")
fwp.add_argument("title", help="Substring to search for (case-insensitive)")
wfwp = sub.add_parser("wait-for-window", help="Poll until a window with given title appears")
wfwp.add_argument("session_id")
wfwp.add_argument("title", help="Substring to wait for (case-insensitive)")
wfwp.add_argument("--timeout", type=int, default=30, metavar="SECONDS", help="Max wait time (default: 30s)")
wfwp.set_defaults(func=cmd_wait_for_window)
pp = sub.add_parser("prompt", help="Show a MessageBox asking the user to do something manually")
pp.add_argument("session_id")
pp.add_argument("message", help="What to ask the user (e.g. 'Please click Save, then OK')")